La lecture en ligne est gratuite
Le téléchargement nécessite un accès à la bibliothèque YouScribe
Tout savoir sur nos offres
Télécharger Lire

Evaluation of training for the unemployed [Elektronische Ressource] : new evidence on effect heterogeneity, dropouts, and program duration / Marie Elina Waller

259 pages
Evaluation of Training for theUnemployed - New Evidence on EfiectHeterogeneity, Dropouts, andProgram DurationInauguraldissertation zur Erlangung des akademischenGrades eines Doktors der Wirtschaftswissenschaften derUniversit˜at MannheimMarie Elina Wallervorgelegt im Juli 2009Referent: Prof. Bernd Fitzenberger, Ph.D.Korreferent: PD Dr. Friedhelm PfeifierAbteilungssprecher: Prof. Tom Krebs, Ph.D.Tag der mundlic˜ hen Prufung:˜ 14. Dezember 2009AcknowledgementsFirst and foremost, I am deeply grateful to my supervisor Bernd Fitzenberger. Ini-tially as a graduate student in his classes at CDSE Mannheim and later on workingin his team at Goethe-University Frankfurt, at ZEW and flnally at Albert-Ludwigs-University Freiburg, I beneflted enormously from his guidance, his support, his sug-gestions and his advice on countless occasions. Our numerous discussions formedmy understanding of empirical work and labor economics.I would like to thank Friedhelm Pfeifier for helpful comments and for being thesecond supervisor of my dissertation. Also, I would like to thank my co-authorsfor our productive cooperation. To the Center for Doctoral Studies in Economics(CDSE) at the University of Mannheim I am thankful for the opportunity to beingpartoftheirgraduateprogramIstronglybenefltedfrom. Theempiricalworkinthisdissertation is based on data from the Institute of Employment Research (IAB).
Voir plus Voir moins

Evaluation of Training for the
Unemployed - New Evidence on Efiect
Heterogeneity, Dropouts, and
Program Duration
Inauguraldissertation zur Erlangung des akademischen
Grades eines Doktors der Wirtschaftswissenschaften der
Universit˜at Mannheim
Marie Elina Waller
vorgelegt im Juli 2009Referent: Prof. Bernd Fitzenberger, Ph.D.
Korreferent: PD Dr. Friedhelm Pfeifier
Abteilungssprecher: Prof. Tom Krebs, Ph.D.
Tag der mundlic˜ hen Prufung:˜ 14. Dezember 2009Acknowledgements
First and foremost, I am deeply grateful to my supervisor Bernd Fitzenberger. Ini-
tially as a graduate student in his classes at CDSE Mannheim and later on working
in his team at Goethe-University Frankfurt, at ZEW and flnally at Albert-Ludwigs-
University Freiburg, I beneflted enormously from his guidance, his support, his sug-
gestions and his advice on countless occasions. Our numerous discussions formed
my understanding of empirical work and labor economics.
I would like to thank Friedhelm Pfeifier for helpful comments and for being the
second supervisor of my dissertation. Also, I would like to thank my co-authors
for our productive cooperation. To the Center for Doctoral Studies in Economics
(CDSE) at the University of Mannheim I am thankful for the opportunity to being
partoftheirgraduateprogramIstronglybenefltedfrom. Theempiricalworkinthis
dissertation is based on data from the Institute of Employment Research (IAB). I
am grateful to the IAB, especially to Stefan Bender, for making the data available.
FinancialsupportbytheIABandtheDeutscheForschungsgemeinschaftisgratefully
acknowledged.
My time at CDSE Mannheim was an outstanding experience to me due to the
exceptional friendship and mutual support I experienced from the co-students of
my year. I would like to thank in particular Stefanie Brilon, Hans-Martin von
Gaudecker, Alia Gizatulina, Florian Muller,˜ and Melanie Schienle.
I am grateful to the people I met at ZEW for their suggestions and to my colleagues
at Frankfurt, namely Martin Biewen, Karsten Kohn, Aderonke Osikominu, and
Robert V˜olter, from whom I learned a lot, in particular on empirical research. Also
at Freiburg, I beneflted from working in a friendly and lively environment.
To my parents, I want to express my strong gratefulness. Among countless other
things,Iamgratefultothemforhavingpreparedmesowellforuniversityeducation.
Dennis lived with me the whole experience of this dissertation - my deep gratitude
goes out to him.Contents
General Introduction 1
1 Which Program for Whom? Evidence on the Comparative Efiec-
tiveness of Public Sponsored Training Programs in Germany 11
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.2 Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.3 Training as Part of Active Labor Market Policy . . . . . . . . . . . . 17
1.4 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
1.4.1 Integrated Biographies Sample . . . . . . . . . . . . . . . . . . 21
1.4.2 Evaluation Sample and Training Programs . . . . . . . . . . . 24
1.5 Econometric Implementation . . . . . . . . . . . . . . . . . . . . . . . 26
1.5.1 Multiple Treatments in a Dynamic Context . . . . . . . . . . 26
1.5.2 Speciflcation of the Propensity Scores . . . . . . . . . . . . . . 30
1.5.3 Estimating Efiect Heterogeneity . . . . . . . . . . . . . . . . . 32
1.6 Empirical Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
1.6.1 Training vs. ‘Waiting’ . . . . . . . . . . . . . . . . . . . . . . 33
1.6.2 Pairwise Evaluation of Training Programs . . . . . . . . . . . 41
1.6.3 Cumulated Efiects . . . . . . . . . . . . . . . . . . . . . . . . 46
1.6.4 Efiect Heterogeneity . . . . . . . . . . . . . . . . . . . . . . . 49
1.7 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
2 On the Importance of Correcting Reported End Dates of Labor
Market Programs 73
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
2.2 End Dates of Labor Market Programs in the IEBS. . . . . . . . . . . 75
2.2.1 Data Set . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
2.2.2 Relevance of End Dates . . . . . . . . . . . . . . . . . . . . . 77
I2.2.3 Error-Proneness of End Dates for Labor Market Programs . . 78
2.3 Empirical Approach: The Example of Further Training . . . . . . . . 80
2.3.1 Plausibility Checks . . . . . . . . . . . . . . . . . . . . . . . . 80
2.3.2 Three Procedures to Deal with Error-prone End Dates . . . . 81
2.3.3 Treatment and Sample . . . . . . . . . . . . . . . . . . . . . . 85
2.4 Sensitivity Analysis I: Frameworks with a Simple Treatment Variable 87
2.4.1 Impact on Employment Rates of Participants . . . . . . . . . 87
2.4.2 Impact on Treatment Efiects Using Matching . . . . . . . . . 89
2.5 Sensitivity Analysis II: Framework with Time-varying Treatment
Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
2.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
3 D¶eja? Vu? Short-Term Training in Germany 1980-1992 and 2000-
2003 101
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
3.2 Institutional Background . . . . . . . . . . . . . . . . . . . . . . . . 106
3.3 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
3.3.1 Administrative Data Sets Used . . . . . . . . . . . . . . . . . 109
3.3.2 Sample Selection . . . . . . . . . . . . . . . . . . . . . . . . . 111
3.4 Evaluation Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
3.5 Empirical Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
3.5.1 Estimation of the Propensity Scores . . . . . . . . . . . . . . . 117
3.5.2 Estimated Treatment Efiects . . . . . . . . . . . . . . . . . . . 118
3.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
Additional Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
4 Many Dropouts? Never Mind! - Employment Prospects of
Dropouts from Training Programs 157
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
4.2 Identiflcation of Dropouts of Further Training Programs in the IEBS 161
4.2.1 The Integrated Employment Biographies Sample. . . . . . . . 161
4.2.2 Sample and Further Training Programs . . . . . . . . . . . . . 162
4.2.3 Identiflcation of Dropouts in the Data . . . . . . . . . . . . . 163
4.3 Descriptive Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . 166
4.3.1 Occurrence of Dropout . . . . . . . . . . . . . . . . . . . . . . 166
II4.3.2 Employment Rates and Employment Stability . . . . . . . . . 168
4.4 Joint Estimation of Dropout and Employment: Does Dropout Harm
in the Long Run? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
4.4.1 The Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
4.4.2 MCMC Estimation . . . . . . . . . . . . . . . . . . . . . . . . 177
4.4.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
4.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
5 The Heterogeneous Efiects of Training Incidence and Duration on
Labor Market Transitions 197
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
5.2 Institutional Background and Data . . . . . . . . . . . . . . . . . . . 204
5.2.1 Training in Germany . . . . . . . . . . . . . . . . . . . . . . . 204
5.2.2 Constructing a Panel Data Set. . . . . . . . . . . . . . . . . . 206
5.2.3 Descriptive Analysis . . . . . . . . . . . . . . . . . . . . . . . 208
5.3 Evaluation Framework . . . . . . . . . . . . . . . . . . . . . . . . . . 211
5.3.1 Estimation Approach . . . . . . . . . . . . . . . . . . . . . . 211
5.3.2 MCMC Estimation of a Random Efiects Probit Model . . . . 215
5.3.3 Estimation of the Treatment Efiects of Interest . . . . . . . . 216
5.4 Estimation Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
5.4.1 Model Fit and Selection on Unobservables . . . . . . . . . . . 219
5.4.2 Classical Treatment Efiect on Employment Probability . . . . 220
5.4.3 Training versus Waiting . . . . . . . . . . . . . . . . . . . . . 222
5.4.4 Variation in Planned Training Duration. . . . . . . . . . . . . 223
5.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225
Appendix A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
Appendix B . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232
Appendix C . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233
List of Tables V
List of Figures VIII
IIIIVGeneral Introduction
The efiectiveness of training programs for the unemployed, as well as other parts
of active labor market policies (ALMP), has been an important topic in the inter-
national literature in labor economics over the last decades. There are still many
open issues, but fast progress has been made in recent years. This was enabled
by advancements in microeconometric methods as well as by a rising interest of
policymakers in econometric evaluations which enhanced the composition of data
sets based on process generated data made available to researchers (see for example
Card, Kluve, and Weber (2009)).
This statement applies not only internationally, but also to the particular situation
in Germany. The Federal Employment O–ce of Germany is ofiering a wide range
of ALMP and in particular difierent types of training, ranging from short programs
which essentially aim at activating the unemployed, to further training
intending to considerably increase the human capital of participants, to very long
retraining programs which lead to a degree in a new profession. Each year there are
more than one million entries into public sponsored training programs. Thus, it is
of strong interest whether these programs are efiective, for whom and under which
circumstances. Furthermore, because of the wide range of large-scale programs
ofiered, the German ALMP constitute a fruitful fleld of study to labor economists
interested in analyzing difierent aspects of the efiectiveness of ALMP.
When I started working on the dissertation project at the beginning of 2005, com-
paratively little was known about the efiectiveness of these programs. Most of the
existing studies were based on small data sets and relatively restrictive econometric
methods. Inmostcasesitwasnotpossibletodistinguishbetweendifierentgroupsof
participants or difierent types of training programs. These studies mostly found no
1oronlyverysmallpositiveefiectsoftraining Whilealmostalloftheearly
studies had to rely on data sufiering from major constraints (i.e. small sample size,
poor deflnition of program participation, no possibility to distinguish between dif-
ferent programs), in 2004 flrst results had become known from extensive projects in
Germany which aimed at producing and utilizing large and rich research-data from
2difierent administrative data sources of the Federal Employment Agency. One of
1For a literature survey on the evaluation of German training programs until the beginning of
2005, see Schneider et al. (2006). Heckman et al. (1999) survey the early international literature.
2FirststudiesusingadministrativedatafromtheFederalEmploymentAgencyareFitzenberger,
1these projects produced the flrst versions of the so-called Integrated Employment
Biographies Sample (IEBS). The IEBS is a large and rich data set combining data
from four administrative data sources. These data allow to study various questions
related to unemployment and ALMP. Its availability enhanced research on these
3topics and also made the empirical work of this dissertation possible.
Recent methodological progress strongly enhanced the quality of microeconometric
evaluation studies as well. Let me name three exemplary topics which have been
very in uential for applied work, including this dissertation. First, the work of
Imbens (2000) and Lechner (2001) on pairwise comparison of multiple treatments
provides the framework to compare difierent programs, i.e. to answer the question
what would have happened to a participant if he or she had been assigned to a
program that difiers from the one he or she is actually assigned to. Second, much
progress has been made with regard to program evaluation in dynamic settings. In
adynamicsetting, theso-calledtiming-of-eventsbecomesimportantasdiscussedby
Fredriksson and Johansson (2003) and Sianesi (2004). Static treatment evaluations
implicitlyconditiononfutureoutcomesleadingtopossiblybiased treatmentefiects.
The nontreated individuals in the data might be observed as nontreated because
their treatment starts after the end of the observation period or because they exit
unemployment before treatment starts (Fredriksson and Johansson (2003)). Third,
Abbring and van den Berg (2003) propose an estimation strategy which explicitly
usesthetiming-of-eventsinadynamicsettingtoidentifythetreatmentefiectapply-
ingacontinuousdurationmodel. Apartfromtheseinnovationswhichrelatedirectly
to program evaluation, this dissertation makes use of relatively recent progress in
Markov Chain Monte Carlo (MCMC) methods, a technique which has been ad-
vanced by Bayesian statisticians in particular in the 1990s (see Chib (2001) for an
overview).
The flve chapters of this dissertation represent flve stand-alone research papers.
They cover various aspects of the evaluation of training programs like efiect hetero-
geneity, comparison of difierent program types, data quality, dynamic selection into
programsandoutofprograms,occurrenceandemploymentperspectivesofdropouts,
Osikominu, and V˜olter (2008), Fitzenberger and Speckesser (2007), Hujer, Thomsen, and Zeiss
(2006), Lechner, Miquel, and Wunsch (2007, 2009). Klose and Bender (2000) use part of these
data even at an earlier time.
3MeanwhilethereareacoupleofotherstudiesusingtheIEBSandfocussingontraining: Kluve,
Schneider, Uhlendorfi, and Zhao (2007), Lechner and Wunsch (2006), Osikominu (2008), Rinne,
Sc and Uhlendorfi (2007), Rinne, Uhlendorfi, and Zhao (2008), Schneider and Uhlendorfi
(2006), and Wunsch and Lechner (2008).
2

Un pour Un
Permettre à tous d'accéder à la lecture
Pour chaque accès à la bibliothèque, YouScribe donne un accès à une personne dans le besoin